The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing remote notification payloads (userInfo), which are untrusted external data sources.
Ingestion points: Data enters the application through application(_:didReceiveRemoteNotification:) and userNotificationCenter(_:didReceive:) callbacks (detailed in references/apns-lifecycle.md and references/notification-runtime.md).
Boundary markers: The skill recommends using a PayloadValidator and checking authorization status before processing content.
Capability inventory: The skill includes patterns for network operations (token upload), file system writes (storing notification attachments), and navigation routing.
Sanitization: The instructions explicitly warn to "Treat payload data as untrusted" and require validation of identity and authorization against current app state before acting on the data.
[EXTERNAL_DOWNLOADS]: The skill contains multiple links to documentation hosted on sosumi.ai. This domain is a well-known placeholder/parody domain used in the Apple developer community for documentation mirrors and tutorials. These links are informational and do not facilitate the download or execution of untrusted scripts.

push-notifications